On Monday, 24 June 2013 at 15:29:55 UTC, deadalnix wrote:
On Monday, 24 June 2013 at 14:36:03 UTC, bearophile wrote:
qznc:
To make dmd 1% faster, calcHash must be reduced to 69% run
time. Does not look possible.
I think the main problem to solve is to reduce the amount of
memory used during the compilation.
Yes, my profiling show that were I spend the most time is
swapping. This have the effect of making every single part of
DMD slower, and have potential for a 10x speedup (easy !).
That, and it means dmd has better chances of finishing its
compile before crashing (*cough*algorithm.d -unittest*cough*)
